基于国产AI芯片的目标检测算法优化与部署

doi:10.3969/j.issn.1006-2475.2025.01.005

摘要/Abstract

摘要： 目前各类神经网络已经逐步在社会的各个方面得到广泛应用，神经网络模型的性能很大程度上取决于其训练策略的优劣，其落地部署也离不开相应硬件平台的支持，而为了保障当前形势下我国的信息安全和电子信息产业发展，相关国产AI芯片的替代迫在眉睫。本文以国产AI芯片替代为出发点，基于全爱QA-200RC开发套件，探究在国产平台上的神经网络算法部署流程，针对特定任务需求进行YOLOv6神经网络训练和主机程序的优化，在通过摄像头进行实时检测的情况下，实现对火箭残骸的目标检测，帧率达30 FPS，mAP_0.5为90.1%，功耗为8.1 W，满足了边缘平台上完成目标检测任务的需求，对促进国产芯片在相关领域的应用具有一定作用。

关键词: 目标检测, 神经网络, 模型训练, AI芯片, 国产化, 硬件平台, Python

Abstract: At present， various types of neural networks have gradually been widely applied in all aspects of society. The performance of neural network models largely depends on the quality of their training strategies， and their deployment cannot be separated from the support of corresponding hardware platforms. In order to ensure the information security and development of the electronic information industry in China under the current situation， it is urgent to replace relevant domestic AI chips. Taking the replacement of domestic AI chips as the starting point， this article explores the deployment process of neural network algorithms on domestic platforms based on the Quanai QA-200RC development kit. The improvement of YOLOv6 neural network training and host program optimization are carried out according to specific task requirements. With real-time detection through cameras， target detection of rocket debris is achieved， the frame rate is 30 FPS， the mAP_0.5 is 90.1%， and the power consumption is 8.1 W， which meets the requirements for completing object detection tasks on edge platforms and is helpful for promoting the application of domestic chips in related fields.

Key words: target detection, neural network, model training, AI chip, localization, hardware platform, Python

中图分类号:

TP391

陈思贇1, 马怀波2, 张华君2, 兰子柠2, 陈文鑫2, 胡杰1, 常胜1. 基于国产AI芯片的目标检测算法优化与部署[J]. 计算机与现代化, 2025, 0(01): 25-29.

CHEN Siyun1, MA Huaibo2, ZHANG Huajun2, LAN Zining2, CHEN Wenxin2 , HU Jie1, CHANG Sheng1. Optimization and Deployment of Object Detection Algorithm Based on Domestic AI Chips[J]. Computer and Modernization, 2025, 0(01): 25-29.

参考文献

［1］赵永强，饶元，董世鹏，等. 深度学习目标检测方法综述［J］. 中国图象图形学报， 2020，25（4）:629-654.
［2］ HONG S， KIM H. An integrated GPU power and performance model［C］// Proceedings of the 37th Annual International Symposium on Computer Architecture. ACM， 2010:280-289.
［3］孙明，陈昕. 面向卷积神经网络的硬件加速器设计方法［J］. 计算机工程与应用， 2021，57（13）:77-84.
［4］谭洪贺，余凯. 端侧AI芯片的挑战和展望［J］. 人工智能， 2018（2）:113-121.
［5］ AMARA A， AMIEL F， EA T. FPGA vs. ASIC for low power applications［J］. Microelectronics Journal， 2006，37（8）:669-677.
［6］ PUJOL R， TABANI H， KOSMIDIS L， et al. Generating and exploiting deep learning variants to increase heterogeneous resource utilization in the Nvidia Xavier［C］// The 31st Euromicro Conference on Real-time Systems （ECRTS 2019）. ECRTS， 2019：23：1-23：23.
［7］乐学. 谷歌推出定制化机器学习芯片速度是传统GPU的15到30倍［J］. 信息与电脑（理论版）， 2017（8）:7-8.
［8］ KIMM H， PAIK I， KIMM H. Performance comparision of TPU， GPU， CPU on Google colaboratory over distributed deep learning［C］// The 2021 IEEE 14th International Symposium on Embedded Multicore/Many-core Systems-on-Chip （MCSoC）. IEEE， 2021:312-319.
［9］ WANG Y， WEI G Y， BROOKS D. A systematic methodology for analysis of deep learning hardware and software platforms［J］. Proceedings of Machine Learning and Systems， 2020，2:30-43.
［10］ SENGUPTA J， KUBENDRAN R， NEFTCI E， et al. High-speed， real-time， spike-based object tracking and path prediction on Google edge TPU［C］// The 2020 2nd IEEE International Conference on Artificial Intelligence Circuits and Systems （AICAS）. IEEE， 2020:134-135.
［11］ NAFFZIGER S， BECK N， BURD T， et al. Pioneering chiplet technology and design for the AMD EPYC™ and Ryzen™ processor families: Industrial product［C］// The 2021 ACM/IEEE 48th Annual International Symposium on Computer Architecture （ISCA）. IEEE， 2021:57-70.
［12］ YE W， HU R， ENEV M. Put deep learning to work: Accelerate deep learning through Amazon SageMaker and ML services［C］// Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining. ACM， 2020:3496-3496.
［13］ SU C L， LAI W C， ZHANG Y K， et al. Artificial intelligence design on embedded board with edge computing for vehicle applications［C］// 2020 IEEE 3rd International Conference on Artificial Intelligence and Knowledge Engineering （AIKE）. IEEE， 2020:130-133.
［14］尹首一，郭珩，魏少军. 人工智能芯片发展的现状及趋势［J］. 科技导报， 2018，36（17）:45-51.
［15］冯昭奎. 中美芯片之争:现实、逻辑与思考［J］. 亚太安全与海洋研究， 2023（2）:18-36.
［16］ GIRSHICK R， DONAHUE J， DARRELL T， et al. Rich feature hierarchies for accurate object detection and semantic segmentation［C］// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. IEEE， 2014:580-587.
［17］ HE K M， ZHANG X Y， REN S Q，et al. Spatial pyramid pooling in deep convolutional networks for visual recognition［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2015，37（9）:1904-1916.
［18］ GIRSHICK R. Fast R-CNN［C］// Proceedings of the 2015 IEEE International Conference on Computer Vision（ICCV）. IEEE， 2015:1440-1448.
［19］ JIAO L C， ZHANG F， LIU F， et al. A survey of deep learning-based object detection［J］. IEEE Access， 2019，7:128837-128868.
［20］ REDMON J， DIVVALA S， GIRSHICK R， et al. You only look once: Unified， real-time object detection ［C］// The 2016 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. IEEE， 2016:779-788.
［21］ LIU W， ANGUELOV D， ERHAN D， et al. SSD: Single shot MultiBox detector［C］// 14th European Conference on Computer Vision. Springer， 2016:21-37.
［22］李延满，王必恒，赵羚焱. 基于轻量化YOLOv5的安全帽检测［J］. 计算机与现代化， 2023（10）:59-64.
［23］ LI C Y， LI L L， JIANG H L， et al. YOLOv6: A single-stage object detection framework for industrial applications［J］. arXiv reprint arXiv:2209.02976， 2022.
［24］ SAYDIRASULOVICH N S， ABDUSALOMOV A， JAMIL M K， et al. A YOLOv6-based improved fire detection approach for smart city environments［J］. Sensors 2023， 23（6 ）. DOI: 10.3390/s23063161.
［25］ CHEN J L， LIU X. Real-time detection method for infrared pedestrian［C］// The 4th International Conference on Artificial Intelligence and Electromechanical Automation （AIEA 2023）. SPIE， 2023:641-646.
［26］ WU H H， ZHOU Z， FENG M， et al. Real-time single object detection on the UAV［C］// 2019 International Conference on Unmanned Aircraft Systems （ICUAS）. IEEE， 2019:1013-1022.
［27］ ALIBABAEI K， ASSUNÇÃO E， GASPAR P D， et al. Real-time detection of vine trunk for robot localization using deep learning models developed for edge TPU devices［J］. Future Internet， 2022，14（7）. DOI: 10.3390/fi14070199.

[1]	徐胜超, 陈富强 . 基于BP神经网络的食品安全风险预警方法[J]. 计算机与现代化, 2025, 0(01): 20-24.
[2]	何思达, 陈平华. 基于意图的轻量级自注意力序列推荐模型[J]. 计算机与现代化, 2024, 0(12): 1-9.
[3]	张晓东1, 白广芝1, 李敏1, 李昊洋2. 基于经验小波变换的油气井产量预测模型 [J]. 计算机与现代化, 2024, 0(12): 53-58.
[4]	刘宝宝, 杨菁菁, 陶露, 王贺应. 基于注意力的DSMSC的遥感图像场景分类[J]. 计算机与现代化, 2024, 0(12): 72-77.
[5]	陈亮, 李诚, 易伟, 熊伟, 汪晓帆, 唐海东. 基于毫米波雷达与视觉融合的电力现场安全帽佩戴检测[J]. 计算机与现代化, 2024, 0(12): 100-107.
[6]	张宇1, 2, 黎靖1, 2, 马铭1, 2, 王众祥1, 2, 孙妍1, 2. YOLOLW:一个新的轻量级目标检测模型[J]. 计算机与现代化, 2024, 0(11): 91-98.
[7]	马钰, 杨勇, 任鸽, 帕力旦·吐尔逊. 基于GCN和微调BERT的作文自动评分方法[J]. 计算机与现代化, 2024, 0(09): 33-37.
[8]	陈雪松1, 李衡1, 王浩畅2. 结合注意力机制和Mengzi模型的短文本分类[J]. 计算机与现代化, 2024, 0(09): 101-106.
[9]	高帅鹏, 王怡凡. 基于图像的群体情绪识别综述[J]. 计算机与现代化, 2024, 0(08): 98-107.
[10]	来坤. 字符模糊的中文纸质发票文字识别方法[J]. 计算机与现代化, 2024, 0(08): 114-119.
[11]	周宪溪, 牟莉. 基于改进TF-IDF和AGLCNN的新闻长文本分类模型[J]. 计算机与现代化, 2024, 0(08): 120-126.
[12]	杨江1, 孙晓梅1, 许韬2. 基于业务内容构建股票关联关系的股价预测[J]. 计算机与现代化, 2024, 0(07): 21-25.
[13]	刘存莉1, 雷占占2, 郑澳2. 基于循环卷积神经网络的排水管网缺陷检测方法[J]. 计算机与现代化, 2024, 0(07): 26-35.
[14]	武军凯, 茅正冲. 基于CDKF-RBFPID的激光器恒流源控制器[J]. 计算机与现代化, 2024, 0(07): 100-105.
[15]	庞梅, 汪珙, 詹泳, 黄哲法. 基于YOLOv5改进算法的海洋水下垃圾检测方法[J]. 计算机与现代化, 2024, 0(07): 120-126.